Percolation site datasets for Nexus-CAT validation and performance benchmark
收藏资源简介:
This record contains the two benchmark datasets used to characterize the Nexus-CAT clustering and percolation-analysis package (https://github.com/jperradin/nexus). Both archives hold synthetic atomic configurations in extended-XYZ format and are fully reproducible from the generation scripts shipped in the benchmarks/ directory of the source repository. 1. performance_data_generated_20260528.tar.gz (≈12 MB) Synthetic binary A-B simple-cubic lattices of increasing size, used to measure the wall-time and memory scaling of the Nexus-CAT clustering strategies (distance, bond, coordination, shared). 12 extended-XYZ files, benchmark-<N>--<L>.xyz, where <N> is the atom count and <L> the cubic side in reduced units. Sizes range from N = 50 000 (L=50) to N = 600 000 (L = 100). Includes generate_simple_cubic_lattice.py, the deterministic generator (RNG seeded with seed = 1_000_000 * L + N), so the datasets reproduce byte-for-byte. 2. validation_data_generated_20260528.tar.gz (≈2.6 GB) 3D site-percolation configurations on a simple cubic lattice are used to validate Nexus-CAT against the well-known results of the percolation theory universality class. 714 extended-XYZ files organized into one subdirectory per lattice size, L ∈ {20, 25, 30, 35, 40, 45, 50}. percolation_sites_0.3116.xyz - 100 independent frames at p = p_c for each L. percolation_sites_<p>.xyz - sweep over occupation probability p ∈ [0.20, 0.40] in steps of 0.002 for finite-size scaling and critical exponent fits. Each frame is generated from a unique seed: seed = 1_000_000 * L + 1000 * p_index + n (critical point reserved as p_index = 500), fully reproducible with the same Numpy/Numba versions. Reproducibility and usage All files are plain extended-XYZ and can be read by any standard parser (ASE, OVITO, etc.) or directly by Nexus-CAT. The generation scripts, input lists, pre-computed reference results, and the analysis pipeline that produces the published figures are available in the benchmarks/ directory of the Nexus-CAT repository (https://github.com/jperradin/nexus). See benchmarks/README.md for full instructions for regeneration and analysis.



